Indicator Audit · Batch 2

Classic market rules, fourteen sealed audits

We tested ten published rules on the stock and bond markets their authors used: Faber’s 10-month average, Siegel’s 200-day filter, the Halloween effect, the turn of the month, dual momentum, sector momentum, VIX stretches, the golden cross on the Nasdaq-100, 60/40 rebalancing bands and Connors’ RSI(2). The protocol was sealed before we fetched a single price. None of the 14 audits beat its benchmark with statistical support: 0 of 14 survive Holm’s correction, and all 14 grade F.

How it was done

  1. Preregistered, then sealed. One document fixed, for all 14 audits: each rule as published with its source, a grid of commonly tried variants (132 in total), the markets, costs, dates, a holdout from 2018, each benchmark, Holm’s correction across the 14, and a promise to publish every result. It was timestamped by DigiCert and FreeTSA (RFC 3161) on 2026-09-25 at 20:11:44 UTC, before any data were fetched.
  2. Run once each under our sealed rubric v1.1. Every report cites the seal.
  3. Batch test. Each audit’s Hansen SPA p-value (the best variant of its family against the benchmark) is corrected with Holm’s method. An audit “survives” only if its adjusted p is 0.05 or less.

What we found

Every SPA p-value was at least 0.68. Two defaults finished slightly ahead of their benchmark, but neither was significant:

  • The golden cross on QQQ compounded at 12.0% a year against 9.4% for holding. Its whole advantage came from sitting out the 2000–02 crash; from 2003 it trailed.
  • 12-1 sector momentum compounded at 9.3% against 8.5% for SPY, with a PBO of 0.99 and a weaker holdout.

The calendar rules, the Halloween effect and the turn of the month, trailed holding by wide margins.

Out of the market, every rule earned 0% in cash: no T-bill series was used, and we said so before running. That works against the timing rules, but it is small next to most of the shortfalls. These are statistics on past data only, not advice, and not a verdict on these ideas in general.

Equities and equity ETFs

RuleMarketVariantsGradeCompound return a year vs benchmarkExcess SharpeDSRPBOHoldout excess Sharpe (in → out)Holm pSurvives
Faber 10-month SMA timingSPY4F9.5% vs 10.9%−0.160.1270.64−0.02 → −0.541.000no
Faber 10-month SMA timingEFA4F7.7% vs 7.6%−0.070.2990.65−0.04 → −0.171.000no
200-day moving-average filterSPY8F8.7% vs 10.8%−0.200.0350.14−0.15 → −0.361.000no
Halloween indicator / Sell in MaySPY6F6.9% vs 10.9%−0.350.0020.13−0.23 → −0.761.000no
Halloween indicator / Sell in MayEFA6F6.5% vs 6.6%−0.080.1470.56−0.07 → −0.141.000no
Turn-of-the-month effectSPY6F3.2% vs 10.9%−0.510.0010.33−0.44 → −0.701.000no
Sector momentum rotation (12-1 month)9 sector SPDRs12F9.3% vs 8.5%+0.070.4910.99+0.16 → −0.121.000no
VIX stretches (Connors & Alvarez)SPY (VIX signal)18F2.5% vs 10.9%−0.540.0000.00−0.40 → −0.981.000no
Golden cross (50/200-day)QQQ24F12.0% vs 9.4%+0.030.2870.50+0.08 → −0.181.000no
Connors RSI(2) with exit variantsQQQ12F2.8% vs 8.9%−0.350.0110.01−0.17 → −0.791.000no
Connors RSI(2) with exit variantsIWM12F2.2% vs 8.9%−0.390.0120.87−0.40 → −0.371.000no

Multi-asset and bonds

RuleMarketVariantsGradeCompound return a year vs benchmarkExcess SharpeDSRPBOHoldout excess Sharpe (in → out)Holm pSurvives
Faber 10-month SMA timingIEF4F1.0% vs 3.3%−0.490.0030.20−0.89 → +0.021.000no
Antonacci dual momentum (Global Equities Momentum)SPY/EFA/AGG6F9.7% vs 11.2%−0.150.1590.53+0.02 → −0.591.000no
60/40 rebalancing bandsSPY/IEF 60/4010F8.6% vs 9.5%−0.380.0150.64+0.12 → −0.721.000no

Sources. Faber (2007), J. Wealth Management; Siegel (2002), Stocks for the Long Run; Bouman & Jacobsen (2002), American Economic Review; Lakonishok & Smidt (1988), RFS, and McConnell & Xu (2008), FAJ; Antonacci (2014), Dual Momentum Investing; Moskowitz & Grinblatt (1999), J. Finance, and Faber (2010); Connors & Alvarez (2009), Short Term Trading Strategies That Work; Brock, Lakonishok & LeBaron (1992), J. Finance; Jaconetti, Kinniry & Zilbering (2010), Vanguard. Full citations are in each report and in the preregistration.

Seal. Preregistration SHA-256 ff8cace63cad05977286a8a9a33dbbc88071c2f12a1d4a0d78f87f8d33f98250; seal record 35f52f363e7cb128df873803d8e10268977adb2cf06204cb6dfd1b2e57f18b11. The preregistration, receipts and code are in the open-source repository (samples/batch2/).

Disclosures. Single-asset rules are compared with buy-and-hold of the same ETF. Dual momentum, sector rotation and the VIX rule are compared with SPY, and the 60/40 bands with an unrebalanced 60/40. The VIX index is used as a signal only. We had already seen SPY results from earlier audits. There were no deviations from the preregistered protocol.

Prices from Yahoo Finance’s public chart endpoint; we publish derived statistics only, never raw prices; no redistribution.

These audits analyze published rules on historical data for illustration. They are not recommendations to trade any of these strategies, and a grade is not a prediction.